Quick way to identify process(es) eating up swap

[mkinney]

Hello,

I have a system with swap at 100%. I'm looking for a quick way to identify if it's a runaway process(es) killing swap. Another way of putting it is that I'm trying to determine root cause for the swap utilized at 100%

Thanks!

 

[mkinney]

FYI, it's a Sun 420r running Solaris 5.7

 

[SamBones]

Keep in mind that on Solaris the directories [tt]/var/run[/tt] and [tt]/tmp[/tt] use swap, so you can look for some very large files in these directories. A big file in [tt]/tmp[/tt] could be your culpret.

If that's not the problem, you can use the command "[tt]prstat -s size[/tt]" to list processes sorted by memory used, or "[tt]prstat -s rss[/tt]" to list the processes by resident memory size.

Also, when you have a suspect process, the command [tt]pmap[/tt] can be used to get details of a processes memory usuage.

Hepe this helps.

 

[kduffin]

It's probably worth noting that /tmp does not have to be swap. Having /tmp be a TMPFS file system is worthwhile for applicatoins that need faster access to those tmp files (such as DBMS systems), but is not a must. I've seen more than one admin with a skewed TMPFS ratio bring his box to its knees.

Cheers,

Keith

 

[drice75]

Keith is right, if you don't need frequent access to /tmp, just change the type of FS in vfstab for /tmp (or even remove it, if your / is big enough).
On my servers, only the Checkpoint Database use /tmpfs, anyway they all got 2 Go of RAM !!! ;o)
See you

 

[blainepruitt]

Not sure if Solaris 2.7 has prstat, if not you can try downloading top from sunfreeware.com. It has information on all the processes on a server, just like prstat, and you can sort by swap.

-BP